> ## Content Index > Fetch the complete content index at: https://www.notatechguy.com/llms.txt > Use this file to discover other available public pages before exploring further. # OlmoEarth v1.3 covers Earth at 10m, no GPU required - URL: https://www.notatechguy.com/olmoearth-v1-3-covers-earth-at-10m-no-gpu-required/ - Published: 2026-10-25T12:25:54.000Z - Updated: 2026-10-25T12:25:54.000Z - Description: AllenAI's OlmoEarth v1.3 releases precomputed satellite embeddings for the entire planet at 10-meter resolution, with fully open weights. - Author: Marcello Babbili - Tags: Technology & AI, AI Chips AllenAI has shipped OlmoEarth v1.3, a satellite embedding model that compresses a year of Earth observation data into a single vector for every 10-meter pixel on the planet [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). Previously, working with OlmoEarth required users to obtain GPU resources and download raw satellite images. The latest version provides precomputed embeddings for the entire globe, accessible via a web browser without any login [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). **My read:** This is the first satellite foundation model I've seen that ships with the entire planet pre-embedded and a browser tool on top. The 20-of-27 benchmark win is impressive on paper, but every number in this release is self-reported, including the PLANTEUR benchmark the team created themselves. I'd want an independent group to reproduce those evals before calling it state-of-the-art. What I find genuinely interesting is the cost trajectory: 53% of v1.2's compute, 18% of v1's. If that holds up, it says something about how quickly satellite model training is maturing. The model behind the embeddings is a ViT-Base with 128 million parameters, paired with a 14.3-million-parameter Perceiver that handles variable-length time series and a tiny linear layer compressing 768-dimensional tokens down to 128 [P²](https://huggingface.co/allenai/OlmoEarth-v1%5F3-Base?ref=notatechguy.com). It reads 12 months of radar from Sentinel-1 and multispectral imagery from Sentinel-2 and Landsat, then outputs one 128-dimensional vector per pixel [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The first 64 dimensions also work as a standalone, smaller embedding [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). ### Self-reported benchmarks show mixed results AllenAI evaluated the 128-dimensional embeddings against two competing models, AlphaEarth Foundations and Tessera v2, across 27 tasks spanning six continents [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The 128-dim version scored highest on 20 of those 27 evaluations. The 64-dim version led on 18, despite being half the size [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). All of those results are self-reported. No independent group has reproduced these evaluations, and the results lack reported error bars [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The evaluation also introduced PLANTEUR, a new crop-type benchmark for tropical farming that the same team built [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). A benchmark created by the model's own developers is a useful signal, not a neutral verdict. The claim that embeddings generalize best to places a classifier has never seen is the team's own characterization, not independently tested [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). ![Benchmark evaluations won out of 27](https://storage.ghost.io/c/6e/89/6e896869-22ef-4281-a213-b4c462c17cff/content/images/2026/10/chart_2c6f7ce1358abb2bbb45.png) ### Browser access replaces GPU provisioning The practical shift is access. Before v1.3, the main path into OlmoEarth was fine-tuning inside OlmoEarth Studio, which required GPU access and enough storage to hold satellite imagery [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The new release puts precomputed annual embeddings on Source Cooperative and wraps them in the OlmoEarth Embeddings Explorer, a browser tool that needs no account and no machine learning background [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The release highlights several challenging geospatial applications, such as identifying smallholder crops in Kenya, spotting brick kilns near Dhaka, evaluating fire damage in Indonesia, and monitoring gold mining activities in the Peruvian Amazon [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). Those are illustrative examples from the release, not independently verified deployments. The team has made the model weights, pretraining data, and code openly accessible [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). As of January 2025, the GitHub training repository has accumulated 323 stars [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com)[P³](https://github.com/allenai/olmoearth%5Fpretrain/?ref=notatechguy.com). When applied as a fine-tuning model, v1.3 achieved results on par with v1.2\. It also required less computing power than prior versions, utilizing 53% of v1.2's compute and 18% of v1's, while reducing carbon emissions to 50% and 22% of those earlier iterations [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). For an agritech startup that wants to classify crop types across a region without provisioning GPUs, the precomputed embeddings on Source Cooperative are the first thing to check. Instead of downloading terabytes of raw imagery and spinning up cloud GPUs, an analyst can pull precomputed vectors directly into a local notebook to train a simple classifier [S¹](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com). The model card on Hugging Face lists the architecture and parameter counts [P²](https://huggingface.co/allenai/OlmoEarth-v1%5F3-Base?ref=notatechguy.com). The training code lives on GitHub under the allenai/olmoearth\_pretrain repository [P³](https://github.com/allenai/olmoearth%5Fpretrain/?ref=notatechguy.com). --- *Sources: [S1 — OlmoEarth v1.3: State-of-the-art embeddings for the whole planet](https://huggingface.co/blog/allenai/olmoearth-v1-3?ref=notatechguy.com) · [P2 — allenai/OlmoEarth-v1\_3-Base · Hugging Face](https://huggingface.co/allenai/OlmoEarth-v1%5F3-Base?ref=notatechguy.com) · [P3 — allenai/olmoearth\_pretrain](https://github.com/allenai/olmoearth%5Fpretrain/?ref=notatechguy.com) · [P4 — stactools-packages/sentinel1](https://github.com/stactools-packages/sentinel1/?ref=notatechguy.com) · [P5 — ecovision-uzh/ssl4eco](https://github.com/ecovision-uzh/ssl4eco?ref=notatechguy.com)* --- *Written from 5 sourced items, 5 of them primary.* ## More from Not A Tech Guy - [Herdr: terminal runtime keeps AI coding agents alive, 45k stars](https://www.notatechguy.com/herdr-terminal-runtime-keeps-ai-coding-agents-alive-45k-stars/) - [Anthropic Agent Skills spec hits 26,000 GitHub stars](https://www.notatechguy.com/anthropic-agent-skills-spec-hits-26-000-github-stars/) - [LiteLLM AI gateway hits 61,877 GitHub stars, claims 8ms latency](https://www.notatechguy.com/litellm-ai-gateway-hits-61-877-github-stars-claims-8ms-latency/)